A Model for Abnormal Detection of In-Vehicle CAN Messages Based on Hyperparameter Optimized CNN

Effectively identifying and defending against cyberattacks through intelligent means has become an important research direction for ensuring the safety of intelligent connected vehicles. The paper constructs a novel intrusion detection system framework using CNN, knowledge transfer and model ensemble methods, along with hyperparameter tuning strategies. First, a data transformation model is established to convert CAN message information into images, retaining the key information and features from the original messages while providing good visualization effects and compatibility, thereby facilitating the identification of different network attack patterns. Secondly, a novel intrusion detection system framework is built using CNN, knowledge transfer and model ensemble methods, along with hyperparameter tuning strategies, which can effectively detect various attack features targeting in-vehicle networks. Finally, the effectiveness of the framework is verified using benchmark datasets, and the detection rate data is analyzed alongside other cutting-edge frameworks, showing that this approach delivers outstanding performance and is feasible for practical application.